ORIGINAL RESEARCH ARTICLEpublished: 23 August 2013

doi: 10.3389/fmicb.2013.00244

Transcriptional regulation of the carbohydrate utilizationnetwork in Thermotoga maritimaDmitry A. Rodionov1,2*, Irina A. Rodionova1, Xiaoqing Li1, Dmitry A. Ravcheev 1,2,

Yekaterina Tarasova 3, Vasiliy A. Portnoy 3, Karsten Zengler3 and Andrei L. Osterman 1

1 Sanford-Burnham Medical Research Institute, La Jolla, CA, USA2 A. A. Kharkevich Institute for Information Transmission Problems, Russian Academy of Sciences, Moscow, Russia3 Department of Bioengineering, University of California San Diego, La Jolla, CA, USA

Edited by:

Aindrila Mukhopadhyay, LawrenceBerkeley National Laboratory, USA

Reviewed by:

Dong-Woo Lee, Kyungpook NationalUniversity, South KoreaAindrila Mukhopadhyay, LawrenceBerkeley National Laboratory, USA

*Correspondence:

Dmitry A. Rodionov,Sanford-Burnham Medical ResearchInstitute, 10901 North Torrey PinesRoad, La Jolla, CA 92037, USAe-mail: rodionov@burnham.org

Hyperthermophilic bacteria from the Thermotogales lineage can produce hydrogen byfermenting a wide range of carbohydrates. Previous experimental studies identified a largefraction of genes committed to carbohydrate degradation and utilization in the modelbacterium Thermotoga maritima. Knowledge of these genes enabled comprehensivereconstruction of biochemical pathways comprising the carbohydrate utilization network.However, transcriptional factors (TFs) and regulatory mechanisms driving this networkremained largely unknown. Here, we used an integrated approach based on comparativeanalysis of genomic and transcriptomic data for the reconstruction of the carbohydrateutilization regulatory networks in 11 Thermotogales genomes. We identified DNA-bindingmotifs and regulons for 19 orthologous TFs in the Thermotogales. The inferred regulatorynetwork in T. maritima contains 181 genes encoding TFs, sugar catabolic enzymesand ABC-family transporters. In contrast to many previously described bacteria, atranscriptional regulation strategy of Thermotoga does not employ global regulatoryfactors. The reconstructed regulatory network in T. maritima was validated by geneexpression profiling on a panel of mono- and disaccharides and by in vitro DNA-bindingassays. The observed upregulation of genes involved in catabolism of pectin, trehalose,cellobiose, arabinose, rhamnose, xylose, glucose, galactose, and ribose showed a strongcorrelation with the UxaR, TreR, BglR, CelR, AraR, RhaR, XylR, GluR, GalR, and RbsRregulons. Ultimately, this study elucidated the transcriptional regulatory network andmechanisms controlling expression of carbohydrate utilization genes in T. maritima. Inaddition to improving the functional annotations of associated transporters and catabolicenzymes, this research provides novel insights into the evolution of regulatory networksin Thermotogales.

Keywords: carbohydrate metabolism, transcriptional regulation, regulon, comparative genomics, Thermotoga

INTRODUCTIONCarbohydrates constitute the most abundant single class oforganic substances found in nature. Plant cell walls consti-tute ∼70% of the worldwide biomass production by landplants, but only ∼2% of this biomass is currently utilized byhumans (Pauly and Keegstra, 2010). Carbohydrate composi-tion of plant biomass is highly diverse and differs significantlybetween plants. A limited number of mono—or disaccharidescompose a majority of plant biopolymers. Cellulose, hemicellu-loses, and pectins are major polysaccharides of the plant cell wall.Hemicelluloses are characterized by a large diversity of build-ing blocks including pentoses (xylose, arabinose) and hexoses(glucose, mannose, galactose, and rhamnose, and uronic acids).Pectins are composed of galacturonate and rhamnose residueswith various branching side chains. Microbial degradation ofplant cell wall polysaccharides and their conversion to biofu-els is a vital objective for society and requires novel efficienttechnologies.

Sugar utilization pathways are major feed lines of carbon andenergy for central metabolism in a large variety of heterotrophicbacteria. Although these pathways and their transcriptional reg-ulation were extensively studied in model bacteria, projectionof this knowledge to diverse bacteria is a major challenge dueto chemical diversity of carbohydrates and a matching variabil-ity of microbial sugar utilization genes, pathways, and regulons.This variability includes alternative biochemical routes, non-orthologous gene replacements, and functionally heterogeneousfamilies of paralogs. Due to this complexity, genomic annotationsof sugar utilization genes derived solely from sequence similarityanalysis are often imprecise and incomplete, especially for dis-tantly related bacteria. Metabolic reconstruction based on a com-bination of various types of genomic context analysis (Yang et al.,2006; Rodionov et al., 2007; Leyn et al., 2012), primarily operonsand regulons, allows us to substantially improve the quality offunctional annotations and predictions, enabling more accuratemetabolic modeling. In this study, we applied an integrative

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 1

Rodionov et al. Regulation of Thermotoga sugar catabolism

approach to simultaneous genomics-based reconstruction of bothmetabolic and associated regulatory networks to study carbohy-drate utilization networks in a hyperthermophilic marine bac-terium from the phylogenetically deep-branching Thermotogalesgroup.

Thermotoga spp. are anaerobic fermentative bacteria that areable to grow on various simple and complex carbohydratesincluding glucose, starch, cellobiose, xylan, and pectin while pro-ducing hydrogen, carbon dioxide, and acetate (Chhabra et al.,2002; Kluskens et al., 2003; Conners et al., 2006). T. maritimaMSB8, a model bacterium in the Thermotoga group, was isolatedfrom geothermally heated marine sediments of Volcano Island inItaly (Huber et al., 1986). A closely related bacterium, T. neapoli-tana, was isolated from a submarine thermal vent near Lucrino,Bay of Naples, Italy (Jannasch et al., 1988). Other Thermotogaleshave a broad geographic distribution (see Figure 1) includ-ing hydrothermal vents in the Azores [Themotoga sp. RQ-2,(Swithers et al., 2011)], a sulfate-reducing bioreactor in Europe[T. lettingae, (Balk et al., 2002)], a hot spring in New Zealand[Fervidobacterium nodosum, (Patel et al., 1985)], and an offshoreoil reservoir in Japan [T. naphthophila, T. petrophila, (Takahataet al., 2000)].

The complete genome sequence of T. maritima revealed alarge proportion of genes (10–15%) involved in carbohydratemetabolism (Nelson et al., 1999). A number of T. maritimaand T. neapolitana enzymes that can degrade β-glucan, hemi-celluloses, and pectin have been studied by functional genomicsor biochemistry (Conners et al., 2006). A three-dimensionalreconstruction of the central metabolic network of T. maritimaincludes 478 enzymes, 120 of which were determined experimen-tally (Zhang et al., 2009). Still, many metabolic gaps remained insugar catabolic networks of T. maritima. We recently investigatedsome of those gaps using a combination of bioinformatics andexperimental techniques (Rodionova et al., 2012a,b, 2013).

The comprehensive genome-scale metabolic modeling of T.maritima also requires an understanding of mechanisms, compo-nents, and behavior of the transcriptional regulatory machinery.The integration of genome-scale metabolic and regulatory modelssignificantly improves growth phenotype predictions and allowsfor the interpretation of systems biology datasets (Faria et al.,2013). Functional genomics approaches were previously usedto study global gene expression in response to growth of T.maritima on sugars. The transcriptional responses of T. mar-itima to various mono- and polysaccharides identified subsets

FIGURE 1 | Properties of 11 Thermotogales species analyzed in

this study. (A) The geographic isolation sites for theThermotogales strains. (B) Distribution of sugar utilizationpathways and cognate transcriptional regulators. The presence oforthologous genes encoding regulators and associated sugar

catabolic pathways is shown by “+” and colored circles,respectively. The phylogenetic species tree was constructed usingthe concatenated alignment of 78 universal bacterial proteins inthe MicrobesOnline database (http://www.microbesonline.org/cgi-bin/

speciesTree.cgi).

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 2

Rodionov et al. Regulation of Thermotoga sugar catabolism

of genes that are coordinately regulated in response to spe-cific carbohydrates (Chhabra et al., 2003; Nguyen et al., 2004;Conners et al., 2005; Frock et al., 2012). However, the molecularmechanisms underlying these transcriptional responses remainunclear.

To infer sugar-responsive transcriptional regulons inThermotoga, we started from the initial metabolic reconstruction,applying a combination of comparative genomics-based methodsof regulatory reconstruction (Rodionov, 2007). Previously, weapplied these methods to reconstruct a large number of metabolicregulons in diverse taxonomic groups of bacteria (Rodionovet al., 2011; Leyn et al., 2013; Ravcheev et al., 2013). Integration ofmetabolic and regulatory reconstructions is particularly efficientfor elucidation of carbohydrate utilization networks in a groupof taxonomically related bacteria, as recently demonstrated bythe comparative genomic study of Shewanella spp. (Rodionovet al., 2010). In previous studies, we also combined genomics-based inferences with experimental elucidation to identify andcharacterize several novel regulons in Thermotoga, includingRex controlling the central carbon metabolism and hydrogenproduction (Ravcheev et al., 2012) and seven ROK-familytranscription factors controlling the sugar utilization pathways(Kazanov et al., 2013).

Here, we extended this analysis toward the entire tran-scriptional network for sugar catabolism in T. maritima com-pared to 10 additional Thermotogales species with completelysequenced genomes. The reconstructed networks allowed usto improve gene annotations, refine associated pathways, andidentify novel, yet uncharacterized, enzymes and transporterstentatively implicated in the sugar utilization machinery. Someof the critical inferences about newly identified regulons wereexperimentally verified. A comparison with global transcriptomicdata obtained for T. maritima grown on different carbohydratesprovided additional validation and enrichment of the genomicreconstruction.

RESULTSGENOMIC RECONSTRUCTION OF SUGAR CATABOLIC REGULONS INTHERMOTOGALESInitially, we used the subsystems-based bioinformatics workflow(see Methods for details) to identify the repertoire of known andputative genes involved in carbohydrate utilization in T. maritima(Table S1 in Supplementary Material). Functional annotationof the identified genes was conducted using the comparativegenomic analysis of predicted gene operons and regulons inthe Thermotogales that was combined with homology searchesand metabolic reconstruction. The identified set of ∼240 genesencodes the components of at least 19 peripheral sugar utiliza-tion pathways as well as the central carbohydrate metabolism(Table 1). In T. maritima, the predicted sugar catabolome uti-lizes 127 enzymes, 22 transport systems (encoded by 86 genes), 18DNA-binding transcriptional factors, (TFs), and 1 RNA-bindingregulator (GlpP).

Next, we identified orthologs for the T. maritima sugarcatabolic, transport, and regulatory genes in 10 additionalThermotogales species with complete genomes (Table S2 inSupplementary Material). Orthologs of the predicted sugar

regulators are unevenly distributed among the 11 analyzedgenomes (Figure 1). We note that the presence of the orthologouscognate TFs is not ubiquitous for some sugar catabolic pathways.The most conserved TFs in the Thermotogales include TreR andGlpR (present in 10 genomes); GalR, GlpP, and UgtR (present in9 genomes), as well as CelR and ChiR (present in 8 genomes).Three regulators (AraR, BglR, and XylR) were found only in6 Thermotoga genomes, whereas four regulators (UxaR, KdgR,ManR, and UctR) are conserved in 5 closely-related Thermotogaspp. Finally, GloR, GluR, InoR, and RhaR have orthologs in asmall subset of Thermotoga spp.

To infer novel TF regulons, we applied the comparativegenomics–based approach implemented in the RegPredict Webserver (see Methods). As a result, we revealed DNA motifs andreconstructed regulons for all 18 DNA-binding TFs associatedwith sugar catabolism in T. maritima. In addition, we inferreda regulon for the predicted fructose regulator FruR, which ispresent in two Thermotoga species but not in T. maritima. Theglycerol regulon operated by the RNA-binding antiterminatorprotein GlpP was not studied by comparative genomics.

The obtained TF-binding motifs and respective sets of co-regulated operons in T. maritima are illustrated in Figure 2.Regulon content for orthologous TFs in other Thermotogalesgenomes is available in the RegPrecise database (Novichkov et al.,2010a) and in Table S2 in Supplementary Material. Metaboliccontent of 15 functionally annotated regulons is visualized onthe integrative T. maritima sugar catabolic network (Figure 3).Below, we describe the reconstructed transcriptional regulatorynetwork for sugar catabolism in the Thermotogales in more detail.

The xylose-responsive regulon XylR in T. maritima containsseven operons organized into five chromosomal gene clustersthat encode xylose catabolic enzymes (XylA, XylB), a xylose-specific ABC transporter, a set of hydrolytic enzymes specificfor xylan and xylan oligosaccharides, and two xylan-inducedABC transporters. XylR belongs to the ROK (repressor, openreading frame, kinase) protein family. All six predicted XylR-binding sites have been experimentally validated in vitro inour previous study of ROK-family regulators (Kazanov et al.,2013). The XylR regulon is well conserved in five Thermotogaspp., whereas T. lettingae has a reduced size of an orthologousregulon.

The predicted regulator KdgR in T. maritima (IclR proteinfamily) is homologous to a repressor of the 2-keto-3-deoxy-gluconate (KDG) metabolism and pectin utilization in enter-obacteria (Rodionov et al., 2004). However, in Thermotoga spp.,the reconstructed KdgR regulon does not include the KDGmetabolism genes (kdgK, kdgA) but is predicted to control theuronate isomerase uxaC, encoding a bifunctional enzyme in thecatabolism of glucuronate and galacturonate. Additional candi-date KdgR-binding sites were identified in the promoter regionsof two XylR-regulated gene clusters, xynB/xlo and xynA/xtp,that are induced during the growth of T. maritima on xylan.Glucuronoxylan is a linear polymer of β-linked xylose residueswith many of the xylose units substituted with glucuronateresidues. The xtp gene cluster encodes α-glucuronidase AguA thatcan release glucuronate residues from glucuronoxylan. Thus, wepropose that the two Thermotoga loci under dual regulation by

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 3

Rodionov et al. Regulation of Thermotoga sugar catabolism

Table 1 | Distribution of the carbohydrate utilization genes and regulators in T. maritima.

Sugar utilization Abbrev. Regulators Transporters Enzymes TOTAL TOTAL genes in

pathway genes regulons

Central carbohydrate metabolism CCM – – 18 18 –

Arabinose, arabinosides Ara AraR AraEFG 6 11 10

β-glucosides Bgl BglR BglEFGKL, 2 9 9

Cellobiose Cel CelR CelEFGKL 7 13 13

Xyloglucan oligosaccharides Glo GloR GloEFGKL 3 9 9

Chitobiose, chitin Chi ChiR ChiEFG 3 7 7

Fructose Fru FruR* FruAB* 2* 7* 6*

Galactose, galactosides Gal GalR LtpEFGKL, GanEFG 7 16 15

Glycerol Glp GlpP GlpABC, GlpF 2 7 –

Digalacturonate, pectin Uxa UxaR AguEFG 15 19 16

Glucuronate Kdg KdgR – 1 3 3

Xylose, xylan Xyl XylR XylEFK, XloEFGKL, XtpEFGKL, XtpN 12 27 27

Inositol Ino IolR InoEFGK 6 11 6

Maltose, maltodextrins Mal – MalEFG 8 15 –

Ribose Rbs RbsR RbsABC 6 12 12

Glucose Glu GluR GluEFK - 4 4

Trehalose Tre TreR TreEFG 1 5 5

Mannose, mannosides Man ManR MtpEFGKL 7 13 11

Rhamnose, rhamnose oligosaccharides Rha RhaR RtpEFGKL 8 14 14

Arabitol, mannitol Pol – – 3 3 –

Hypothetical sugar utilization HSU UgtR, UctR, UgpR UgtEFGK, UctMPQ, UgpEFG 11 24 20

Total number: 19 86 127 241 181

The details of all functional assignments summarized in this table are provided in Table S1. Asterisks indicate that FruR regulator and fructose utilization genes are

absent from T. maritima and thus were not counted toward the total numbers.

XylR and KdgR are involved in the xylan and glucuronoxylanutilization.

The predicted regulator UxaR (GntR family) was attributed toa conserved 17-nt DNA motif found in multiple copies withinthe pectin/galacturonate utilization gene loci in five Thermotogagenomes (Rodionova et al., 2012a). The UxaR-regulated pectincatabolic locus in T. maritima includes pectate lyase pelA, digalac-turonate ABC transporter aguEFG, exo-polygalacturonase pelB,and the gnd, uxaE, uxaD, gntE, and gntK genes required forgalacturonate catabolism. Interestingly, the T. maritima aguEFGtransporter is substituted by other UxaR-regulated transportsystems (named uxaPQM and uxaXYZ) in T. petrophila, T.naphtophila, and T. neapolitana. A conserved UxaR binding sitewas identified upstream of the kdgA-kdgK-uxuB-uxuA operonencoding enzymes of the glucuronate utilization pathway. Theseresults suggest that the pectin/galacturonate and glucuronatecatabolic genes are co-regulated by the transcription factorUxaR in Thermotoga spp. The UxaR regulon predicted by insilico genomic analysis was experimentally validated in T. mar-itima (see Experimental validation of the inferred regulons inT. maritima).

The mannose-responsive regulon ManR (ROK family)includes two operons encoding a set of mannan and mannosideshydrolytic enzymes and a mannan-induced ABC transporter.Both candidate binding sites of ManR and its mannose effectorhave been previously experimentally tested in T. maritima(Kazanov et al., 2013). The ManR-dependent regulation ofthe manBDCR operon is conserved in other Thermotoga spp.,

whereas the ManR-regulated mtp-cel5 operon is conserved onlyin the RQ-2 strain.

The predicted cellobiose regulator CelR (LacI family) has anextensive regulon including 22 genes from six genomic loci inT. maritima. The conserved part of the CelR regulon (present infive Thermotoga genomes) includes glucan and β-glucoside degra-dation enzymes, cellobiose phosphorylase, and a barley glucan–induced ABC transporter. The latter transporter, termed celE-FGKL, is also present and regulated by orthologous CelR regula-tors in three other Thermotogales genomes (Figure 1). In T. mar-itima, the reconstructed CelR regulon includes an additional ABCtransporter (gloEFGKL) encoded within the predicted xyloglucanutilization gene cluster. The LacI-family transcription factor GloRwas predicted to function as a local regulator of the glo locus.The common intergenic region of gloR and gloE genes containsdistinct operator sites for CelR and GloR regulators. The BglRregulator (ROK family) controls another β-glucosides utilizationoperon encoding laminarinase, β-glucosidase, and cellobiose-and laminaribiose-specific ABC transporters in Thermotoga spp.A BglR-binding site was experimentally validated, and it wasfound that the BglR regulator from T. maritima responds to twoeffectors, cellobiose and glucose (Kazanov et al., 2013).

The ROK-family transcription factors GluR and TreR in T.maritima control the glucose and trehalose utilization path-ways (Kazanov et al., 2013). The trehalose-responsive regulatorTreR co-regulates the trehalose ABC transporter treEFG and thetrehalose synthase-like gene treT, whereas GluR controls theglucose ABC transporter gluEFK in response to glucose. The

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 4

Rodionov et al. Regulation of Thermotoga sugar catabolism

FIGURE 2 | Genomic context of 15 sugar catabolic regulons in T.

maritima. Validated and predicted effectors of TFs are listed in red andblack, respectively. TF binding sites and downstream regulated genes areshown by circles and arrows, respectively. Validated sites are labeled withred check marks. Sequence logos representing the consensus binding sitemotifs were built using all candidate sites in the Thermotogales genomes.

Genes encoding transcriptional regulators and components of sugartransporters are shown in black and red, respectively, whereas the genesencoding the secreted and intracellular sugar catabolic enzymes are ingreen and yellow, respectively. Hypothetical genes are in gray. Genes thatwere upregulated on specific sugars (listed in the last column) have theirnames highlighted in magenta.

glucose regulator has an additional binding site upstream ofthe treEFG operon, which is conserved in five closely relatedThermotoga genomes (Figure 4). These results suggest that thetrehalose transporter is under dual control of both regulators. TheTreR regulon is conserved in all studied Thermotogales genomesexcept for Petrotoga mobilis.

Two other ROK-family regulators, IolR and ChiR, control theinositol and chitobiose catabolic operons, respectively. The lat-ter operon includes the predicted chitobiose transporter chiEFG,the chitobiose hydrolase cbsA, and the N-acetylglucosamine(GlcNAc) catabolic genes nagAB. Single DNA binding sites ofboth IolR and ChiR regulators were experimentally confirmed

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 5

Rodionov et al. Regulation of Thermotoga sugar catabolism

FIGURE 3 | Integrative reconstruction of sugar utilization pathways and

regulons in T. maritima. Diagram includes 190 carbohydrate utilizationgenes encoding catabolic enzymes (solid arrows), transporters (dottedarrows), and transcriptional factors (in ovals of different colors). Genes

involved in the reconstructed regulatory network (161 genes regulated by 15TFs) are highlighted by a matching background color. Three hypothetical sugarutilization pathways and the respective regulons (UgtR, UctR, UgpR) are notshown.

(Kazanov et al., 2013). The specific DNA-binding ability of theChiR regulator was abolished by a disaccharide chitobiose butnot monosaccharide GlcNAc. However, the IolR regulator didnot respond to any tested sugar, including glucose, inositol, andinositol phosphate (Kazanov et al., 2013). The IolR regulon isconserved in four Thermotoga genomes, whereas the ChiR reg-ulon is present in all studied Thermotoga and Thermosipho spp.(Figure 1).

The predicted regulator AraR (GntR family) is homologousto the L-arabinose-responsive repressor from Bacillus subtilis(Rodionov et al., 2001). The reconstructed AraR regulon infive Thermotoga species includes predicted ABC transportersand hydrolytic enzymes for utilization of arabinose oligosac-charides and a set of L-arabinose catabolic enzymes. The pre-dicted ribose regulator RbsR (LacI family) is encoded within theribose utilization operon. The conserved part of the RbsR reg-ulon in eight Thermotogales genomes includes a ribose-specific

ABC transporter and a ribokinase. In T. maritima, T. neapoli-tana, and T. lettingae, the ribose operon also includes genesencoding two predicted D-arabinose catabolic enzymes andtransketolase.

The predicted rhamnose regulator RhaR (DeoR family) wasidentified within the rhamnose utilization gene cluster in T. mar-tima and three other Thermotoga spp. The reconstructed RhaRregulon includes a rhamnose-induced ABC transporter, severalglycoside hydrolases, and a set of rhamnose catabolic enzymes.The predicted galactose regulator GalR (LacI family) controls thegalactan/galactoside utilization pathways in nine Thermotogalesgenomes. In T. maritima, the GalR regulon includes two galactosecatabolic enzymes, two β-galactosidases, one galactan hydrolase,and two lactose-inducible ABC transporters. The predicted fruc-tose utilization regulon FruR (DeoR family) was found in onlytwo Thermotoga spp. (T. naphthophila and RQ-2). Finally, threeother TF regulons reconstructed in Thermotoga genomes control

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 6

Rodionov et al. Regulation of Thermotoga sugar catabolism

FIGURE 4 | Genomic context and conservation of trehalose and

glucose utilization genes and regulons in Thermotogales.

yet uncharacterized sugar catabolic pathways (UctR/TM0326,UgtR/TM1228, and UgpR/TM1856).

EXPERIMENTAL VALIDATION OF THE INFERRED REGULONS INT. MARITIMAWe further assessed the reconstructed sugar catabolic regulons byperforming genome-wide transcriptional profiling of T. maritimacells grown on various mono- and disaccharides. First, the abil-ity of T. maritima to grow on 17 different carbon sources wastested in minimal media with a 10-mM concentration of the par-ticular carbohydrate in 500-ml bottles under anoxic conditionsat 80◦C. Growth phenotype testing revealed that T. maritima isable to grow on the monosaccharides rhamnose, galactose, glu-cose, fructose, arabinose, xylose, ribose, and mannose and thedisaccharides trehalose, cellobiose, and maltose as a sole carbonand energy source (Figure 5). In contrast, N-acetylglucosamine,gluconate, glucuronate, galacturonate, mannitol, sorbitol, andinositol did not support the growth of T. maritima and thus werenot used for further transcriptional analysis. The obtained phe-notypes are in agreement with the previous studies that havedemonstrated growth of T. maritima on simple sugars such as glu-cose, mannose, ribose, arabinose, xylose, and rhamnose (Connerset al., 2005), as well as on the disaccharides maltose and lactose(Nguyen et al., 2004).

Using high-density oligonucleotide tiling arrays, we analyzedchanges in transcription levels of T. maritima cells grown ondifferent carbon sources (see Methods). The obtained sugar-specific transcriptomes were compared with ribose and other

FIGURE 5 | Maximum cell concentration for T. maritima grown on

various carbon sources. Cell density was determined by optical densitymeasured at 600 nm (OD600) after 24 h of cultivation. These values wereconverted to cell counts using the following equation [cells/mL] =OD600/3.58 × 10−10, which was determined from measures of OD600taken at numerous time points in the growth curve and correlated with flowcytometric cell counts.

sugars to calculate the differential expression of all genes from reg-ulons reconstructed in T. maritima (Table S3 in SupplementaryMaterial). The obtained gene induction patterns for xylose, cel-lobiose, trehalose, glucose, arabinose, ribose, rhamnose, andgalactose correlate with the reconstructed regulons and metaboliceffectors for the XylR, BglR, CelR, TreR, GluR, AraR, RbsR, RhaR,and GalR regulators. In addition, the UxaR regulon involving thepectin and galacturonate genes was experimentally validated in T.maritima by both the transcriptomic approach and the in vitroDNA-binding assays with the recombinant UxaR protein. Thegenes/operons from these regulons are presented in Figure 2 anddiscussed below with respect to their differential expression andphysiological function in T. maritima.

XylR regulonTranscription levels of all 27 genes from the seven operons reg-ulated by the xylose-responsive transcription factor XylR weresignificantly elevated in T. maritima grown on xylose com-pared to cells grown on ribose. The highest fold changes ofgene expression (∼10- to 20-fold) were observed for two xylanutilization gene clusters, xynB<>xloEFGKL-xyl3-axeA and xynA-cenC<>xtpFGKLE-aguE. The xylose catabolic gene cluster xylR-adhB-xylFUEKB demonstrated modest upregulation on xylose(∼2- to 6-fold), whereas the xylose isomerase xylA had approxi-mately 10-fold higher transcript levels. The xylan oligosaccharideutilization operon xtpN-bgaL was induced approximately 6-fold.In vitro, the XylR regulator responded to 0.02 mM of xylose andalso to >0.2 mM of glucose (Kazanov et al., 2013). However,current in vivo experiments have not identified any significantinduction of the XylR-regulated genes on glucose, suggesting thatxylose, rather than glucose, is a physiological inducer of the XylRregulon in T. maritima.

BglR and CelR regulonsGenes from the BglR-controlled bglREFGKL-TM0026-bglB-lamAoperon were 4- to 6-fold upregulated during growth on cel-lobiose (β-glucosyl-1,4-glucose) and 3- to 11-fold induced by

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 7

Rodionov et al. Regulation of Thermotoga sugar catabolism

glucose. The obtained gene expression profiles for the BglR reg-ulon are corroborated by the previous in vitro DNA-bindingassays with the BglR repressor that was affected by both cel-lobiose and glucose, although the glucose disaccharide was amore effective inducer (Kazanov et al., 2013). All 13 genes thatbelong to five operons from the predicted CelR regulon werehighly upregulated during growth on cellobiose but remaineduninduced during growth on glucose. The cbpA and celQ genesshowed the highest induction by cellobiose (more than 10-fold),whereas the responses of the celEFGKLR, cel12AB, and TM0312-13 operons to cellobiose were more moderate (5- to 9-fold). Incontrast, the putative xyloglucan oligosaccharide utilization genecluster gloR<>gloEFGKL-cel74-fucAI, which was predicted as apart of the CelR regulon, was uninduced on cellobiose or othersugars tested. The latter fact can be explained by transcriptionalrepression of the xlo gene locus by another LacI-family predictedregulator, GloR, that has a yet-unknown sugar effector. Basedon these results, we concluded that cellobiose is a physiologicaleffector that induces the CelR regulon in T. maritima.

TreR and GluR regulonsThe TreR-regulated treTR operon encoding a putative trehaloseutilization enzyme and the trehalose-responsive repressor washighly upregulated in T. maritima cells grown on trehalose(Figure 6A). Although the DNA microarray hybridization con-firmed the in vivo induction of treTR by trehalose, similar analysisof the treEFG and gluEFK genes was not possible due to theabsence of the respective oligonucleotide probes. The T. mar-itima MSB8 isolate that was originally sequenced in 1999 [TIGRgenomovar, (Nelson et al., 1999)] and that was used for microar-ray gene probe design lacks the 8870-bp DNA region between thegenes TM1846 (cbpA) and TM1847 (gluR). This DNA region ispresent in two other T. maritima MSB8 isolates that have recentlybeen resequenced [genomovars DSM 3109 (Boucher and Noll,2011) and ATCC (Latif et al., 2013)]. The new region includesseven genes including bglA, a member of the CelR-regulated cbpA-bglA operon, and the treEFG and gluEFK genes that are regulatedby TreR and GluR (Kazanov et al., 2013) (Figure 6C). To val-idate sugar-specific induction of these two operons in vivo, weperformed real-time reverse transcription PCR (RT-PCR) withprobes designed for treE, gluE, gluR, and bglA (used as a nega-tive control) (Figure 6B). All three genes tested had demonstratedelevated expression levels in T. maritima cells grown on eitherglucose or trehalose compared to the cells grown on ribose. Thehighest fold changes in gene expression were observed for treEand gluE in the glucose-grown cells, whereas in the trehalose-grown cells only the treE gene was highly upregulated. Theseresults confirm the dual control of the trehalose transportertreEFG by both TreR and GluR regulators in response to trehaloseand glucose, respectively. Also, the RT-PCR results confirm thatthe glucose transporter gluEFGR operon is induced by glucoseusing the glucose-responsive repressor GluR.

AraR regulonAll genes from the predicted AraR regulon that are organized inthree transcriptional units were upregulated in T. martima cellsgrown on arabinose compared to cells grown on ribose. The

FIGURE 6 | Differential expression of trehalose and glucose utilization

genes in T. maritima. (A) Gene expression fold changes for the treTR andgluR genes for the cells grown on glucose (Glu) and trehalose (Tre) vs. ribose(Rib), as determined by whole-genome microarrays. (B) Gene expression foldchanges for the treE, gluE, and gluR genes as determined by RT-PCR. ThebglA gene was used as a negative control. Normalized transcript levels withrespect to the TM0688 (gap) gene are provided in Figure S1 inSupplementary Material. The mean of biological duplicate measurementswas used. (C) Genomic clusters of genes involved in the trehalose (red) andglucose (blue) utilizations.

TM0280-abfA and araA genes demonstrated the highest induc-tion levels (61- to 161-fold), the arabinoside transporter genesaraEFG were modestly induced (9- to 16-fold), whereas expres-sion of the arabinose catabolic genes araMDBW was increasedonly 2- to 6-fold. Large differences in the induction levels betweenthe TM0280-abfA and araMDBW genes can be attributed tothe putative transcriptional terminator upstream of araM (Latifet al., 2013). These results suggest that arabinose is a physiologi-cal inducer for the AraR regulon. This hypothesis is supported bythe fact that AraR orthologs in Clostridium acetobutylicum and B.subtilis function as arabinose-responsive repressors (Zhang et al.,2012).

RbsR regulonTranscription of all genes from the ribose utilization operon washighly upregulated in T. maritima cells grown on ribose comparedto cells grown on glucose (14- to 92-fold). Unexpectedly, theribose operon genes were almost equally highly induced in cellsgrown on xylose. The ribose operon is predicted to be controlledby the LacI-family regulator RbsR, whose molecular effector is notyet experimentally determined. The unusual induction patternsfor ribose utilization genes may be explained by the dual speci-ficity of the RbsR repressor to ribose and xylose. Alternatively,since the xylose utilization feeds the pentose-phosphate pathway,it is possible that ribose 5-phosphate, an intermediate of this path-way, serves as a physiological effector of the RbsR regulon in T.maritima.

RhaR regulonThe rhamnose oligosaccharide utilization gene cluster in T. mar-itima includes 14 genes organized in three transcriptional unitsthat are predicted to be regulated by RhaR. Transcription of allthree RhaR-regulated operons was highly elevated in T. mar-itima cells grown on rhamnose compared to cells grown on

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 8

Rodionov et al. Regulation of Thermotoga sugar catabolism

ribose. The rhamnose catabolic genes from the rhaMADBCoperon were induced between 27- and 37-fold, the rhaR-agu4Coperon showed ∼12-fold induction, whereas the rtpEFGKL-gusB-TM1061 operon genes demonstrated the gradual decrease in theirinduction level from 60-fold for the first gene to ∼5-fold for thelatter gene. These expression results suggest that the RhaR regula-tor in T. maritima responds to the rhamnose monosaccharide.

GalR regulonThe predicted galactoside utilization regulon GalR in T. maritimaincludes 15 genes organized into two operons, both of which wereupregulated during growth on galactose. Genes encoding puta-tive transporters for galactose oligosaccharide and galactosides(ganEFG and ltpEFGKL), as well as two β-galactosidases (lacA andganA), demonstrated the highest levels of induction on galactosecompared to ribose (8- to 20-fold). The galactose catabolic genes(galT and galK) and the putative galactose repressor gene galRwere modestly induced by galactose (∼4-fold). These results arein agreement with the previous microarray expression profilingof T. maritima that have demonstrated induction of all above-mentioned GalR regulon genes in cells grown on lactose but noton maltose (Nguyen et al., 2004).

UxaR regulonThe GntR-type transcriptional regulator UxaR was predictedto bind a conserved DNA motif in regulatory regions of fourpectin and galacturonate utilization operons in T. maritima(Figures 7A,B). The UxaR regulon includes the extracytoplas-mic pectate lyase PelA, the digalacturonate-specific transporterAguEFG, the cytoplasmic polygalacturonase PelB, and the exten-sive set of galacturonate catabolic enzymes (Rodionova et al.,2012a). We performed experimental validation of the predictedUxaR regulon by both in vitro and in vivo approaches. To assessspecific binding of the purified recombinant UxaR protein tothe predicted UxaR operators in T. maritima we used a fluores-cence polarization assay (Figure 7C). The results show that UxaRspecifically binds to the synthetic 27-nt DNA fragments contain-ing UxaR binding sites. All tested DNA fragments demonstratedthe concentration-dependent increase of fluorescence polariza-tion, confirming specific interaction between the regulator andDNA fragments. The apparent Kd values for the UxaR proteininteracting with the tested DNA fragments were in the rangeof 40–80 nM. We also tested the influence of potential sugareffectors on protein-DNA interaction, however, the additionof digalacturonate, galacturonate, glucuronate, or hexuronate

FIGURE 7 | Experimental validation of UxaR regulon in T. maritima. (A)

Non-redundant set of UxaR binding sites identified in T. maritima (TM), T.neapolitana (Tnea), Thermotoga sp. RQ-2 (TRQ2), and T. petrophila (Tpet). (B)

Sequence logo for UxaR binding sites. (C) Fluorescence polarization bindingassay of UxaR with target DNA operators in T. maritima. Increasing protein

concentrations were mixed with 27-bp fluorescence-labeled DNA fragmentsof gene regions containing UxaR-binding sites. As a negative control (N.C.),the T. maritima regulator Rex (TM0169) and DNA fragment of the TM0201gene containing the Rex-binding site were used. (D) Gene expression foldchange for T. maritima grown on pectin vs. ribose.

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 9

Rodionov et al. Regulation of Thermotoga sugar catabolism

catabolic pathway intermediates had no effect on the complexformation (data not shown).

Galacturonate does not support the growth of T. maritimawhen used as a single carbon source (data not shown). Thus, weassessed gene expression in cells grown on pectin, a galacturonate-containing polysaccharide, which can support the growth of T.maritima (Kluskens et al., 2003). The expression of ∼50 sugarcatabolic genes was significantly (>2-fold) upregulated in cellsgrown on pectin compared to cells grown on ribose (Table S3 inSupplementary Material). In particular, the expression of 6 genesfrom the UxaR regulon (aguEFG, pelA, and aldH-pelB) showedmore than 17-fold upregulation (Figure 7D). The extracellularpectate lyase gene pelA was expressed 230-fold higher on pectinwhen compared to ribose. Most of the remaining UxaR-regulatedgenes from the aldH-pelB-gnd-uxaRED-gntEK and kdgAK-uxuBAoperons showed a 2- to 5-fold upregulation on pectin. The agu4A-TM0435 genes located immediately downstream of pelA did notshow any upregulation on pectin, suggesting that they are notco-transcribed with pelA. Accordingly, we found a putative rho-independent transcriptional terminator downstream of pelA. Inaddition to the induction of pectin/galacturonate catabolic genes,three gene clusters encoding the beta-glucoside (TM0024-32),arabinose (TM0276-284), and galactoside (TM1190-1204) utiliza-tion pathways were significantly upregulated in T. maritima cellsgrown on pectin. These genes clusters do not belong to the UxaRregulon but were predicted to be regulated by BglR, AraR, andGalR, respectively. Induction of these three regulons by pectinmay be explained by the presence of glucose, arabinose, andgalactose residues in this polysaccharide.

DISCUSSIONTranscriptional regulation is a highly variable component of car-bohydrate utilization networks. Although our current knowledgeof sugar catabolic regulons in model bacteria such as Escherichiacoli and Bacillus subtilis is nearly comprehensive, the projectionof this knowledge to the sequenced genomes of bacteria fromdistant taxonomic groups is still a challenge. The major difficul-ties for bioinformatics-based propagation of TF regulons includeduplications and losses of TFs and their binding sites, rapiddiversification of specificities of TFs toward DNA sites and sugareffectors, non-orthologous replacements of entire regulatory sys-tems, and frequent horizontal transfers. Comparative genomicanalysis of specific sugar utilization pathways has led to sub-stantial progress in the identification and reconstruction of theircognate TF regulons in diverse bacterial lineages (Yang et al.,2006; Rodionov et al., 2010; Leyn et al., 2012). In this study, weused comparative genomics to reconstruct novel TF regulons forsugar catabolism in hyperthermophilic bacteria from the deep-branching lineage Thermotogales (Figure 1). In T. maritima, amodel bacterium in this lineage, the identified sugar catabolicregulatory network includes 18 TFs, 40 TF binding sites, and163 target genes comprising 15 known and three hypotheticalmetabolic pathways (Table 1; Figures 2, 3). All studied sugar-responsive TFs in T. maritima are encoded within their cognateregulated gene loci, and thus are subject to autoregulation.

All sugar utilization regulons reconstructed in Thermotogalesare controlled by TFs that are homologous to known sugar-related

regulators from six families: DeoR, GntR, IclR, LacI, ROK, andRpiR. For most of these TFs, phylogenetic and genome con-text analysis of distant homologs did not reveal their potentialfunctional orthologs outside the Thermotogales lineage. Thesegenomic observations suggest that the analyzed sugar metabolicregulators have evolved and functionally specialized after the sep-aration of the Thermotogales. Our previous phylogenetic analysissuggests that the expansion of the ROK family represented byseven paralogs in T. maritima was likely due to massive dupli-cations and subsequent functional diversification of regulatorsduring the evolution of Thermotogales (Kazanov et al., 2013).The DeoR-family regulator FruR controlling the fructose utiliza-tion operon in two closely related Thermotoga spp. represents anexceptional case of relatively recent horizontal gene transfer fromthermophilic Clostridia. All genes from the fructose utilizationoperon in Thermotoga spp., including fruR, are highly similar tothe orthologous genes in Caldicellulosiruptor spp. (70–80% iden-tity). We propose that the entire fructose utilization operon waslaterally transferred between these two lineages that likely sharethe same ecological niche.

The observed variations in the distribution of sugar uti-lization regulons among the Thermotogales are also remarkable(Figure 1). Only six regulons are present in at least two differ-ent genera (CelR, ChiR, GalR, GloR, RbsR, and TreR), whereasthe remaining regulons are present only in the Thermotoga genus.Moreover, most of the latter regulons are restricted to the group offive closely related Thermotoga spp, and only three regulons haveorthologs in a more distant genome of T. lettingae. Conservationof all Thermotoga-specific regulons except FruR suggests theirlikely emergence in the common ancestor of this genus. Weobserved several cases of species-specific regulon loss in whichan entire regulon (including all operons from a regulated path-way) is missing only in a single Thermotoga spp., including theinositol and ribose utilization regulons in the RQ-2 strain andthe glucose utilization regulon GluR in T. petrophila. In anothercase, the RhaR regulator was lost in a single strain of Thermotogasp. RQ-2, however, the rhamnose utilization operons are stillretained in the genome. Many other cases of the absence of orthol-ogous regulators for sugar catabolic pathways in the genomesof more distant Thermotogales can be explained by their con-trol by non-orthologous TFs that have yet to be characterized(Figure 1).

Overall, the reconstructed carbohydrate utilization regulatorynetwork in T. maritima contains 18 local TF regulons (FruR isabsent from T. maritima), each controlling between one and sevenoperons. In contrast to other model bacteria (e.g., E. coli andB. subtilis) that employ various transcriptional mechanisms forglobal catabolite repression of sugar metabolism, our genomic-based analysis did not reveal any global regulons for sugar uti-lization genes in Thermotoga. This observation correlates withthe genome-wide transciptome data for T. maritima obtainedin this and previous studies (Chhabra et al., 2003; Frock et al.,2012). With the exception of a fructose-specific system in twoThermotoga strains, the Thermotogales genomes lack orthologs ofsugar phosphotranferase (PTS) systems that are essential compo-nents of global carbon catabolic repression mechanisms in bothGram-negative and Gram-positive bacteria (Deutscher, 2008).

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 10

Rodionov et al. Regulation of Thermotoga sugar catabolism

Thus, bacteria from the Thermotogae phylum appear to usedifferent regulatory strategies for sugar utilization.

Another remarkable feature of sugar catabolic networks inThermotoga is the existence of multiple interconnections betweendistinct sugar regulons. Two xylan catabolic gene loci, xtp andxlo, were found to be controlled by the xylose-responsive reg-ulator XylR and the predicted glucuronate utilization regula-tor KdgR in T. maritima (Figure 2) and in other Thermotogagenomes (Table S2 in Supplementary Material). The first genelocus encodes a secreted endoxylanase, a putative xylan oligosac-charide ABC transporter, and a cytoplasmic glucuronidase, whichmight be involved in the cleavage of glucuronate residues froma xylan-derived oligosaccharide. The second gene locus encodesanother secreted endoxylanase, a xylose oligosaccharide ABCtransporter, and a cytoplasmic xylosidase and is presumablyinvolved in the utilization of xylose-containing oligosaccharides.From transcriptomic data in T. maritima, both XylR/KdgR-regulated gene loci are upregulated during growth on xylose(Tables S3 in Supplementary Material) and xylan (Conners et al.,2005). Another gene locus, glo, encoding hypothetical glucoseoligosaccharide utilization genes, is predicted to be controlledby both the locally encoded regulator GloR and the distal reg-ulator CelR, which also controls five other operons involvedin cellobiose and glucan utilization (Figure 2). Lastly, the tre-halose transporter treEFG in five Thermotoga spp. is controlledby the trehalose- and glucose-responsive regulators TreR andGluR, respectively (Figure 4). We confirmed by RT-PCR that thetreE gene in T. maritima is upregulated on glucose and trehalose(Figure 7). The observed partial overlaps between distinct localregulons in Thermotoga point to the existence of coordinated reg-ulatory responses to particular types of complex polysaccharidesand disaccharides.

We assessed the reconstructed sugar regulatory network inT. maritima by microarray gene expression profiling of cellsgrown on a variety of mono- and disaccharides and pectinas a single carbon source. The obtained gene induction pat-terns demonstrate overall consistency with individual regulonsreconstructed by genomic analysis. In addition, the previouslydetermined effectors of the ROK-family regulators XylR, TreR,GluR, and BglR (Kazanov et al., 2013) were found to be ingood agreement with the in vivo gene expression results forxylose, trehalose, glucose, and cellobiose, respectively (Figure 2).In addition, the reconstructed regulons correlated with previousmicroarray expression studies in T. maritima (Chhabra et al.,2003; Nguyen et al., 2004; Conners et al., 2005). For instance,the predicted galactoside utilization regulon GalR was induced bygalactose in our study and was previously shown to be upregu-lated on lactose, a galactose-glucose disaccharide (Nguyen et al.,2004). The predicted cellobiose and glucan utilization regulonCelR was induced by cellobiose in our study and was previ-ously shown to be upregulated by barley and glucomannan(Conners et al., 2005).

The reconstructed regulatory network allowed us to sug-gest and/or refine specific functional assignments for sugar-specific ABC transporters that constitute the most abundantclass of uptake transporters in sugar utilization pathways ofThermotogales. In T. maritima, the previously uncharacterized

transporter ChiEFG was tentatively assigned chitobiose specificitybased on the chitobiose-responsive regulon ChiR (Figure 3).Similarly, the characterized GluR and TreR regulons and theirrespective effectors allowed us to propose glucose and trehalosespecificities for the GluEFK and TreEFG transporters, respectively,and these functional assignments were recently experimentallyvalidated (Boucher and Noll, 2011). Finally, the reconstructedregulons of other Thermotoga species allowed us to predict sugarspecificities for multiple novel ABC transporters that are non-orthologous to transport systems from T. maritima (Table S1 inSupplementary Material). Interestingly, despite of the absence ofFruR regulon, T. maritima demonstrates some growth on fructoseas a sole carbon source (Figure 5). This observation correlateswith the presence of functional fructokinase TM0296 (Rodionovaet al., 2012b), however, the fructose uptake transporter remainsunknown in T. maritima.

In summary, the comparative genomics-based regulon andpathway reconstruction combined with some experimental dataallowed us to identify the integrated metabolic and regulatorynetwork of sugar utilization in T. maritima and revealed its sub-stantial diversity between different Thermotogales species. Ourresults revealed a high level of consistency between the in sil-ico–predicted TF regulons, the in vitro–determined TF effectorspecificities, and the in vivo–measured gene expression changes.The described integrative genomics–based approach for regulonanalysis may be applied to other yet uncharacterized taxonomicgroups of microorganisms for which multiple closely relatedgenomes are available.

MATERIALS AND METHODSFUNCTIONAL GENE ANNOTATION AND METABOLIC RECONSTRUCTIONTo map carbohydrate utilization genes in the T. maritimagenome, we used the bioinformatic workflow that was previouslyapplied for the analysis of Shewanella genomes (Rodionovet al., 2010). The workflow is based on a collection of manuallycurated subsystems in the SEED genomic database capturinga substantial fraction of known sugar utilization pathwaysprojected across microbial genomes (Overbeek et al., 2005).Using a compilation of ∼500 groups of isofunctional homologsfrom various sugar utilization subsystems in SEED, we performedhomology-based scanning of the T. maritima genome. Based ongenome context analysis of the identified gene candidates, wefinally collected ∼240 genes potentially involved in carbohydrateutilization in T. maritima. We also collected experimentalknowledge and literature references for ∼130 proteins from thecollection including 62 sugar catabolic enzymes, 12 bindingcomponents of ABC transporters, and 6 regulators with estab-lished cognate metabolites (substrates, ligands, or effectors).For the remaining ∼50 T. maritima proteins, the previouslyreported gene expression profiles revealed their upregulationduring the growth on specific poly- or monosaccharides usedas a single carbon source (Nguyen et al., 2004; Conners et al.,2005; Frock et al., 2012). The obtained functional gene anno-tations are captured in the SEED subsystem available onlineat http://pubseed.theseed.org/SubsysEditor.cgi?page=ShowSubsystemandsubsystem=Sugarutilizationin Thermotogales andsummarized in Table S1 in Supplementary Material.

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 11

Rodionov et al. Regulation of Thermotoga sugar catabolism

GENOMIC RECONSTRUCTION OF REGULONSPutative TF regulons were reconstructed using the establishedcomparative genomic approach [reviewed in (Rodionov, 2007)]implemented in the RegPredict Web server (http://regpredict.lbl.gov) (Novichkov et al., 2010b). We started from the genomicidentification of reference sets of genomes that encodes orthologsof each TF and their genome context analysis. Initial train-ing sets of potentially co-regulated genes were collected basedon the gene neighborhood analysis of TF orthologs in theMicrobesOnLine database (http://microbesonline.org/) (Dehalet al., 2010). For de novo identification of a candidate TF-bindingmotif in the training set of potential upstream regions, we usedthe “Discover Profile” procedure implemented in RegPredict.Each identified DNA motif was used to scan the Thermotogalesgenomes to predict novel regulon members. Scores of candidatesites were calculated as the sum of positional nucleotide weights.The score threshold was defined as the lowest score observedin the training set. The conserved regulatory interactions withhigh-scored binding sites that involve target genes involved incarbohydrate metabolism were included in the reconstructed reg-ulons. Candidate sites associated with new members of regulonswere added to the training set, and the respective group profilewas rebuilt to improve search accuracy. Sequence logos for thederived group-specific DNA binding motifs were drawn usingthe WebLogo package (Crooks et al., 2004). The details of allreconstructed regulons are captured and displayed in RegPrecise,a specialized database of bacterial regulons (http://regprecise.lbl.gov) (Novichkov et al., 2010a), as a part of the Thermotogalescollection.

DNA MICROARRAY HYBRIDIZATIONT. maritima MSB8 (ATCC:43589) was grown in 500-ml serumbottles containing 200 ml of minimal medium (Rinker andKelly, 2000) with 10 mM of a carbon source under anoxicconditions at 80◦C. Carbon sources D-glucose, D-galactose,L-rhamnose, L-arabinose, D-xylose, D-ribose, D-mannose, N-acetyl-D-glucosamine, D-gluconate, glucuronate, galacturonate,mannitol, sorbitol, trehalose, cellobiose, maltose, and pectinwere obtained from Sigma (St. Louis, MO). The cell growthwas monitored spectrophometrically at 600 nm. Once the mid-logarithmic phase was reached, growth was stopped with 20 ml ofstop solution composed of 5% Trizol and 95% ethanol (Sigma).RNA isolation was performed using a conventional RNAeasymini kit protocol with DNaseI treatment (Qiagen, Valencia,CA). Total RNA yields were quantified by using a NanoDrop(Thermo Fisher Scientific, Waltham, MA) at a wavelength of260 nm, and RNA quality was checked by measuring the sam-ples’ A260/A280 ratio (>1.8). cDNA synthesis was conductedas following: Amino-allyl cDNAs were reverse transcribed from10 μg of purified total RNA and then labeled with Cy3 monore-active dyes (Amersham, GE HealthCare, UK). Labeled cDNAsamples were fragmented to the 50–300-bp range with DNaseI (EpiCentre Biotechnologies, Madison, WI) and interrogatedwith high-density four-plex oligonucleotide tilling arrays con-sisting of 4 × 71,548 probes of variable length spaced across thewhole T. maritima genome (Roche-NimbleGen, Madison, WI).Hybridization, washing, and scanning were performed according

to the manufacturer’s instructions. Two biological replicates wereutilized for each growth condition. Probe level data were nor-malized using the Robust Multiarray Analysis algorithm with-out background correction as implemented in NimbleScan™2.4 software (Roche-NimbleGen). Transcriptome data have beendeposited in NCBI’s Gene Expression Omnibus and are accessiblethrough GEO Series accession number GSE47615 (http://www.

ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE47615).

REAL TIME PCRIndividual transcript levels were measured for four genes from theATCC-derived strain of T. maritima MSB8: cbpA (Tmari_1862),treE (Tmari_1861), gluE (Tmari_1858), and gluR (Tmari_1855)(Genbank Accession CP004077). Genomic RNA was isolatedfrom cells grown on minimal medium supplied with either tre-halose, D-glucose, or D-ribose and collected at the same opti-cal density using the RNeasy minikit (Qiagen, Valencia, CA).Reverse transcription was performed on 10 μg of total RNA. Thereverse transcription mixture (60 μl) contained 10 μg total RNA,75 μg random primers, 1 × 1st Strand buffer, 10 mM dithiothre-itol, 0.5 mM deoxynucleoside triphosphates, 30 U Superase, and1,500 U Superscript II. The mixture was incubated in a thermo-cycler (Bio-Rad, Hercules, CA) at 25◦C for 10 min, 37◦C for 1 h,and then 42◦C for 1 h. The reaction was followed by incuba-tion at 70◦C for 10 min to inactivate the superscript. The RNAwas then degraded by adding 20 μl of 1 N NaOH and incubat-ing at 65◦C for 30 min. After the incubation, 20 μl of 1 N HClwere added to neutralize the solution. QIAquick PCR purifica-tion kits were used to clean up the cDNA synthesis product.Following the purification, the cDNA was quantified and thendirectly used in quantitative PCR (qPCR) reactions. The 50 μlof qPCR reaction volume contained 25 μl Sybr green tag mas-ter mix (Qiagen), 0.2 μM forward primer, 0.2 μM reverse primer,and cDNA as a template. Each qPCR reaction was run in tripli-cate in the Bio-Rad thermocycler (Bio-Rad, Hercules, CA) withthe following settings: 95◦C for 15 min, 94◦C for 15 s, 52◦Cfor 30 s, and 72◦C for 30 s; the denaturation, annealing, andextension steps were repeated for 40 cycles. In order to deter-mine the binding affinity of each primer set, a standard curvewas calculated for each primer and reaction efficiency obtainedfrom it. Using the standard curve, the relative cDNA quantitywas obtained for each gene by normalizing it to the quantityof gap (TM0688) cDNA in the same sample. The gap gene wasused as an expression control as it demonstrated no differen-tial expression across all gene expression arrays in the presentstudy. Normalized transcript levels are provided in Figure S1 inSupplementary Material.

GENE CLONING AND PROTEIN PURIFICATIONThe E. coli BL21 strains containing pMH2T7-derived plasmid thatharbor the T. maritima TM0439 (uxaR) gene with the arabinosepromoter and the His6 tag were a kind gift from Dr. S. Lesley at theJoint Center for Structural Genomics (La Jolla, CA). RecombinantUxaR protein containing an N-terminal His6 tag was overex-pressed in E. coli in 50 ml volume culture and purified using Ni2+-chelating chromatography, as previously described (Yang et al.,2008). Briefly, cells were grown in TB medium to OD600 = 1.4

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 12

Rodionov et al. Regulation of Thermotoga sugar catabolism

at 37◦C induced by 0.15% L-arabinose and harvested after 3 hshaking at 37◦C. Harvested cells were re-suspended in 20 mMHEPES buffer pH 7 containing 100 mM NaCl, 0.03% Brij 35,and 2 mM β-mercaptoethanol supplemented with 2 mM PMSFand a protease inhibitor cocktail (Sigma-Aldrich). Lysozyme wasadded to 1 mg/ml, and the cells were lyzed by freezing-thawingfollowed by sonication. After centrifugation at 18,000 rpm, theTris-HCl buffer (pH 8) was added to the supernatant (50 mM,final concentration), and it was loaded onto an Ni-NTA acidagarose minicolumn (0.3 mL). After washing with the startingbuffer containing 1 M NaCl and 0.3% Brij-35, bound proteinswere eluted with 0.3 mL of the starting buffer containing 250 mMimidazole and then followed by buffer exchange. Finally, proteinwas suspended in 20 mM HEPES pH7.0 containing 1 mM DTTand 0.5 mM EDTA. Protein size, expression level, distributionbetween soluble and insoluble forms, and extent of purifica-tion were monitored by SDS PAGE. Protein concentration wasdetermined by the Quick Start Bradford Protein Assay kit fromBio-Rad.

DNA BINDING ASSAYSThe interaction of the purified recombinant UxaR protein withtheir cognate DNA-binding sites in T. maritima was assessedusing the fluorescence polarization DNA-binding assay. The DNAoligos containing the predicted binding sites were synthesizedby Integrated DNA Technologies. One strand of oligo was 3′-labeled by a fluorescence label, 6-carboxyfluorescin), whereasthe complementary oligo was unlabeled. The double-strandedlabeled DNA fragments were obtained by annealing the labeledoligonucleotides with unlabeled complementary oligonucleotidesat a 1:10 ratio. The labeled 30-bp DNA fragments (1 nM) were

incubated with increasing concentrations of the purified UxaRprotein (10–500 nM) in a total volume of 100 μl of the bindingbuffer containing 20 mM Tris-HCl (pH 7.5), 100 mM NaCl, and0.3 mg/ml BSA at 37◦C for 30 min. Poly(dI-dC) was added tothe reaction mixture as a non-specific competitor DNA at 1 μgto suppress non-specific binding. The fluorescence-labeled DNAwas detected with the FLA-5100 fluorescent image analyzer. ADNA fragment containing the Rex regulator binding site was usedas a negative control (Ravcheev et al., 2012).

ACKNOWLEDGMENTSWe would like to thank Olga Zagnitko from the Fellowshipfor Interpretation of Genomes (FIG) for help with functionalgene annotations and subsystem analysis in the SEED database.We are grateful to Pavel Novichkov from Lawrence BerkeleyNational Laboratory for help with visualization of regulonsin the RegPrecise database and to Haythem Latif for depo-sition of microarray datasets to GEO. This work was sup-ported by the Office of Science and Office of Biological andEnvironmental Research of the U.S. Department of Energy undercontracts DE-FG02-08ER64686 and DE-SC0004999. Additionalfunding was provided by the Russian Foundation for BasicResearch (12-04-33003) and by National Institute of GeneralMedical Sciences (NIGMS) under Protein Structure Initiative(U54 GM094586-04).

SUPPLEMENTARY MATERIALThe Supplementary Materials for this article can be foundonline at: http://www.frontiersin.org/Microbial_Physiology_and_Metabolism/10.3389/fmicb.2013.00244/abstract

REFERENCESBalk, M., Weijma, J., and Stams,

A. J. (2002). Thermotoga lettingaesp. nov., a novel thermophilic,methanol-degrading bacteriumisolated from a thermophilicanaerobic reactor. Int. J. Syst. Evol.Microbiol. 52, 1361–1368. doi:10.1099/ijs.0.02165-0

Boucher, N., and Noll, K.M. (2011).Ligands of thermophilic ABCtransporters encoded in a newlysequenced genomic region ofThermotoga maritima MSB8screened by differential scan-ning fluorimetry. Appl. Environ.Microbiol. 77, 6395–6399. doi:10.1128/AEM.05418-11

Chhabra, S. R., Shockley, K. R.,Conners, S. B., Scott, K.L.,Wolfinger, R. D., and Kelly, R.M. (2003). Carbohydrate-induceddifferential gene expression pat-terns in the hyperthermophilicbacterium Thermotoga maritima.J. Biol. Chem. 278, 7540–7552. doi:10.1074/jbc.M211748200

Chhabra, S. R., Shockley, K. R., Ward,D. E., and Kelly, R. M. (2002).Regulation of endo-acting glycosyl

hydrolases in the hyperthermophilicbacterium Thermotoga maritimagrown on glucan- and mannan-based polysaccharides. Appl.Environ. Microbiol. 68, 545–554. doi:10.1128/AEM.68.2.545-554.2002

Conners, S. B., Mongodin, E. F.,Johnson, M. R., Montero, C. I.,Nelson, K. E., and Kelly, R. M.(2006). Microbial biochemistry,physiology, and biotechnol-ogy of hyperthermophilicThermotoga species. FEMSMicrobiol. Rev. 30, 872–905. doi:10.1111/j.1574-6976.2006.00039.x

Conners, S. B., Montero, C. I., Comfort,D. A., Shockley, K. R., Johnson, M.R., Chhabra, S. R., et al. (2005).An expression-driven approachto the prediction of carbohydratetransport and utilization regu-lons in the hyperthermophilicbacterium Thermotoga maritima.J. Bacteriol. 187, 7267–7282. doi:10.1128/JB.187.21.7267-7282.2005

Crooks, G. E., Hon, G., Chandonia,J. M., and Brenner, S. E. (2004).WebLogo: a sequence logo genera-tor. Genome Res. 14, 1188–1190. doi:10.1101/gr.849004

Dehal, P. S., Joachimiak, M. P., Price,M. N., Bates, J. T., Baumohl, J.K., Chivian, D., et al. (2010).MicrobesOnline: an integratedportal for comparative andfunctional genomics. NucleicAcids Res. 38, D396–D400. doi:10.1093/nar/gkp919

Deutscher, J. (2008). The mech-anisms of carbon cataboliterepression in bacteria. Curr.Opin. Microbiol. 11, 87–93. doi:10.1016/j.mib.2008.02.007

Faria, J. P., Overbeek, R., Xia, F.,Rocha, M., Rocha, I., and Henry, C.S. (2013). Genome-scale bacterialtranscriptional regulatory networks:reconstruction and integratedanalysis with metabolic models.Brief Bioinform. doi: 10.1093/bib/bbs071. [Epub ahead ofprint].

Frock, A. D., Gray, S. R., and Kelly,R. M. (2012). HyperthermophilicThermotoga species differ withrespect to specific carbohydratetransporters and glycosidehydrolases. Appl. Environ.Microbiol. 78, 1978–1986. doi:10.1128/AEM.07069-11

Huber, R., Langworthy, T. A., Konig, H.,Thomm, M., Woese, C. R., Sleytr,U. B., et al. (1986). Thermotogamaritima sp. nov. represents a newgenus of unique extremely ther-mophilic eubacteria growing up to90◦C. Arch. Microbiol. 144, 324–333.doi: 10.1007/BF00409880

Jannasch, H. W., Huber, R., Belkin,S., and Stetter, K. O. (1988).Thermotoga neapolitana sp. nov.of the extremely thermophiliceubacterial genus Thermotoga.Arch. Microbiol. 150, 103–104. doi:10.1007/BF00409725

Kazanov, M., Li, X., Gelfand, M.,Osterman, A., and Rodionov, D.(2013). Functional diversificationof ROK-family transcriptionalregulators of sugar catabolismin the Thermotogae phylum.Nucleic Acids Res. 41, 790–803. doi:10.1093/nar/gks1184

Kluskens, L. D., Van Alebeek, G. J.,Voragen, A. G., De Vos, W. M., andVan Der Oost, J. (2003). Molecularand biochemical characterization ofthe thermoactive family 1 pectatelyase from the hyperthermophilicbacterium Thermotoga maritima.

www.frontiersin.org August 2013 | Volume 4 | Article 244 | 13

Rodionov et al. Regulation of Thermotoga sugar catabolism

Biochem. J. 370, 651–659. doi:10.1042/BJ20021595

Latif, H., Lerman, J. A., Portnoy, V.A., Tarasova, Y., Nagarajan, H.,Schrimpe-Rutledge, A. C., et al.(2013). The genome organizationof Thermotoga maritima reflects itslifestyle. PLoS Genet. 9:e1003485.doi: 10.1371/journal.pgen.1003485

Leyn, S. A., Gao, F., Yang, C., andRodionov, D. A. (2012). N-acetylgalactosamine utilizationpathway and regulon in proteobac-teria: genomic reconstructionand experimental characteri-zation in Shewanella. J. Biol.Chem. 287, 28047–28056. doi:10.1074/jbc.M112.382333

Leyn, S. A., Kazanov, M. D., Sernova,N. V., Ermakova, E. O., Novichkov,P. S., and Rodionov, D. A. (2013).Genomic reconstruction of tran-scriptional regulatory networkin Bacillus subtilis. J. Bacteriol.195, 2463–2473. doi: 10.1128/JB.00140-13

Nelson, K. E., Clayton, R. A., Gill, S. R.,Gwinn, M. L., Dodson, R. J., Haft,D. H., et al. (1999). Evidence for lat-eral gene transfer between Archaeaand bacteria from genome sequenceof Thermotoga maritima. Nature399, 323–329. doi: 10.1038/20601

Nguyen, T. N., Ejaz, A. D., Brancieri,M. A., Mikula, A. M., Nelson, K.E., Gill, S. R., et al. (2004). Whole-genome expression profiling ofThermotoga maritima in responseto growth on sugars in a chemo-stat. J. Bacteriol. 186, 4824–4828.doi: 10.1128/JB.186.14.4824-4828.2004

Novichkov, P. S., Laikova, O. N.,Novichkova, E. S., Gelfand, M. S.,Arkin, A. P., Dubchak, I., et al.(2010a). RegPrecise: a databaseof curated genomic inferences oftranscriptional regulatory inter-actions in prokaryotes. NucleicAcids Res. 38, D111–D118. doi:10.1093/nar/gkp894

Novichkov, P. S., Rodionov, D. A.,Stavrovskaya, E. D., Novichkova, E.S., Kazakov, A. E., Gelfand, M. S.,et al. (2010b). RegPredict: an inte-grated system for regulon infer-ence in prokaryotes by compar-ative genomics approach. NucleicAcids Res. 38, W299–W307. doi:10.1093/nar/gkq531

Overbeek, R., Begley, T., Butler, R.M., Choudhuri, J. V., Chuang, H.Y., Cohoon, M., et al. (2005). Thesubsystems approach to genomeannotation and its use in theproject to annotate 1000 genomes.

Nucleic Acids Res. 33, 5691–5702.doi: 10.1093/nar/gki866

Patel, B. K. C., Morgan, H. W.,and Daniel, R. M. (1985).Fervidobacterium nodosumgen. nov. and spec. nov., anew chemoorganotrophic, cal-doactive, anaerobic bacterium.Arch. Microbiol. 141, 63–69. doi:10.1007/BF00446741

Pauly, M., and Keegstra, K. (2010).Plant cell wall polymers as pre-cursors for biofuels. Curr. Opin.Plant Biol. 13, 305–312. doi:10.1016/j.pbi.2009.12.009

Ravcheev, D. A., Best, A. A., Sernova,N. V., Kazanov, M. D., Novichkov,P. S., and Rodionov, D. A. (2013).Genomic reconstruction of tran-scriptional regulatory networksin lactic acid bacteria. BMCGenomics 14:94. doi: 10.1186/1471-2164-14-94

Ravcheev, D. A., Li, X., Latif, H.,Zengler, K., Leyn, S. A., Korostelev,Y. D., et al. (2012). Transcriptionalregulation of central carbon andenergy metabolism in bacteria byredox-responsive repressor Rex.J. Bacteriol. 194, 1145–1157. doi:10.1128/JB.06412-11

Rinker, K. D., and Kelly, R. M. (2000).Effect of carbon and nitrogensources on growth dynamics andexopolysaccharide production forthe hyperthermophilic archaeonThermococcus litoralis and bac-terium Thermotoga maritima.Biotechnol. Bioeng. 69, 537–547.

Rodionov, D. A. (2007). Comparativegenomic reconstruction of tran-scriptional regulatory networksin bacteria. Chem. Rev. 107,3467–3497. doi: 10.1021/cr068309+

Rodionov, D. A., Gelfand, M.S., and Hugouvieux-Cotte-Pattat, N. (2004). Comparativegenomics of the KdgR regulonin Erwinia chrysanthemi 3937and other gamma-proteobacteria.Microbiology 150, 3571–3590. doi:10.1099/mic.0.27041-0

Rodionov, D. A., Kurnasov, O. V., Stec,B., Wang, Y., Roberts, M. F., andOsterman, A. L. (2007). Genomicidentification and in vitro reconsti-tution of a complete biosyntheticpathway for the osmolyte di-myo-inositol-phosphate. Proc.Natl. Acad. Sci. U.S.A. 104,4279–4284. doi: 10.1073/pnas.0609279104

Rodionov, D. A., Mironov, A. A.,and Gelfand, M. S. (2001).Transcriptional regulation ofpentose utilisation systems in the

Bacillus/Clostridium group ofbacteria. FEMS Microbiol. Lett.205, 305–314. doi: 10.1111/j.1574-6968.2001.tb10965.x

Rodionov, D. A., Novichkov, P. S.,Stavrovskaya, E. D., Rodionova,I. A., Li, X., Kazanov, M. D.,et al. (2011). Comparativegenomic reconstruction oftranscriptional networks con-trolling central metabolism inthe Shewanella genus. BMCGenomics 12(Suppl 1):S3. doi:10.1186/1471-2164-12-S1-S3

Rodionov, D. A., Yang, C., Li, X.,Rodionova, I. A., Wang, Y.,Obraztsova, A. Y., et al. (2010).Genomic encyclopedia of sugar uti-lization pathways in the Shewanellagenus. BMC Genomics 11:494. doi:10.1186/1471-2164-11-494

Rodionova, I. A., Leyn, S. A., Burkart,M. D., Boucher, N., Noll, K. M.,Osterman, A. L., et al. (2013).Novel inositol catabolic pathwayin Thermotoga maritima. Environ.Microbiol. 15, 2254–2266. doi:10.1111/1462-2920.12096

Rodionova, I. A., Scott, D. A.,Grishin, N. V., Osterman,A. L., and Rodionov, D. A.(2012a). Tagaturonate-fructuronateepimerase UxaE, a novel enzyme inthe hexuronate catabolic networkin Thermotoga maritima. Environ.Microbiol. 14, 2920–2934. doi:10.1111/j.1462-2920.2012.02856.x

Rodionova, I. A., Yang, C., Li,X., Kurnasov, O. V., Best, A.A., Osterman, A. L., et al.(2012b). Diversity and ver-satility of the Thermotogamaritima sugar kinome.J. Bacteriol. 194, 5552–5563.doi: 10.1128/JB.01136-12

Swithers, K. S., Dipippo, J. L., Bruce,D. C., Detter, C., Tapia, R., Han,S., et al. (2011). Genome sequenceof Thermotoga sp. strain RQ2,a hyperthermophilic bacteriumisolated from a geothermallyheated region of the seafloornear Ribeira Quente, the Azores.J. Bacteriol. 193, 5869–5870. doi:10.1128/JB.05923-11

Takahata, Y., Nishijima, M., Hoaki,T., and Maruyama, T. (2000).Distribution and physiologicalcharacteristics of hyperther-mophiles in the Kubiki oilreservoir in Niigata, Japan. Appl.Environ. Microbiol. 66, 73–79. doi:10.1128/AEM.66.1.73-79.2000

Yang, C., Rodionov, D. A., Li, X.,Laikova, O. N., Gelfand, M. S.,Zagnitko, O. P., et al. (2006).

Comparative genomics andexperimental characterizationof N-acetylglucosamine utilizationpathway of Shewanella oneidensis.J. Biol. Chem. 281, 29872–29885.doi: 10.1074/jbc.M605052200

Yang, C., Rodionov, D. A., Rodionova,I. A., Li, X., and Osterman, A.L. (2008). Glycerate 2-kinaseof Thermotoga maritima andgenomic reconstruction ofrelated metabolic pathways.J. Bacteriol. 190, 1773–1782.doi: 10.1128/JB.01469-07

Zhang, L., Leyn, S. A., Gu, Y.,Jiang, W., Rodionov, D. A., andYang, C. (2012). Ribulokinaseand transcriptional regulationof arabinose metabolism inClostridium acetobutylicum.J. Bacteriol. 194, 1055–1064. doi:10.1128/JB.06241-11

Zhang, Y., Thiele, I., Weekes, D., Li,Z., Jaroszewski, L., Ginalski, K.,et al. (2009). Three-dimensionalstructural view of the centralmetabolic network of Thermotogamaritima. Science 325, 1544–1549.doi: 10.1126/science.1174671

Conflict of Interest Statement: Theauthors declare that the researchwas conducted in the absence of anycommercial or financial relationshipsthat could be construed as a potentialconflict of interest.

Received: 13 June 2013; accepted: 31 July2013; published online: 23 August 2013.Citation: Rodionov DA, RodionovaIA, Li X, Ravcheev DA, Tarasova Y,Portnoy VA, Zengler K and OstermanAL (2013) Transcriptional regulation ofthe carbohydrate utilization network inThermotoga maritima. Front. Microbiol.4:244. doi: 10.3389/fmicb.2013.00244This article was submitted to MicrobialPhysiology and Metabolism, a section ofthe journal Frontiers in Microbiology.Copyright © 2013 Rodionov,Rodionova, Li, Ravcheev, Tarasova,Portnoy, Zengler and Osterman. This isan open-access article distributed underthe terms of the Creative CommonsAttribution License (CC BY). The use,distribution or reproduction in otherforums is permitted, provided the orig-inal author(s) or licensor are creditedand that the original publication inthis journal is cited, in accordancewith accepted academic practice. Nouse, distribution or reproduction ispermitted which does not comply withthese terms.

Frontiers in Microbiology | Microbial Physiology and Metabolism August 2013 | Volume 4 | Article 244 | 14